Categories Companies Research

Loading the SOTA2 catalog…

Explore

Categories Companies Research

Contribute

Join SOTA2 Contact

SOTA2

© 2026 SOTA2

Build in Bellevue, USA

WizWand is now SOTA2 Research

Papers Benchmarks Datasets Tasks

ASR Error Correction research benchmarks · SOTA2 Research

ASR Error Correction

Benchmarks

Dataset NameSOTA MethodSort

Dataset NameSOTA methodMetricTrendResultsLast Updated

Common Voice Frisian (test)

8.9WER

27

May 20, 2026

Frisian Offline Data (test)

13.8WER

27

May 20, 2026

5.7WER

18

Feb 26, 2026

CommonVoice (CV) (test)

5.8WER

18

Feb 26, 2026

compositional oracle

10.7WER

8

Feb 26, 2026

compositional oracle

2.6WER

8

Feb 26, 2026

SwitchBoard (test)

compositional oracle

0.042Word Error Rate (WER)

8

Feb 26, 2026

CV-accent (test)

compositional oracle

0.079WER

8

Feb 26, 2026

Tedlium-3 (test)

compositional oracle

0.7WER

8

Feb 26, 2026

compositional oracle

2.8WER

8

Feb 26, 2026

compositional oracle

1.1WER

8

Feb 26, 2026

ASR Error Correction Evaluation Set (test)

16.07WER

6

Feb 26, 2026

Internal Dataset (dev)

10.31WER

6

Feb 26, 2026

Internal Dataset (test)

10.22WER

6

Feb 26, 2026

AISHELL-1 (dev)

3.8WER

6

Feb 26, 2026

AISHELL-1 (test)

4.08WER

6

Feb 26, 2026

31WER

5

Jun 25, 2026

0.39WER

5

Jun 25, 2026

34WER

5

Jun 25, 2026

32WER

5

Jun 25, 2026

Common Voice (Persian) SNR = 10 dB

ELN-conditioned model

28.02WER

4

Feb 26, 2026

Common Voice (Persian) SNR = 5 dB

ELN-conditioned model

32.34WER

4

Feb 26, 2026

Common Voice (Persian) Mixed Noise

ELN-conditioned model

24.84WER

4

Feb 26, 2026

Common Voice Persian (Clean)

Fine-tuned (No ELN)

24.06WER

4

Feb 26, 2026

Error-Aware RAG

18.83WER

3

Jun 25, 2026

Page 1 of 2

PreviousNext

Popular datasets

Common Voice Internal dataset MMLU MedQA FLEURS

Follow for updates

Join SOTA2 Contact